Our friends at APICS Magazine published an e-article awhile back that serves as a timely caution to supply chain managers when it comes to trying to forecast future sales. The article, by Fred Tolbert, CPIM, an APICS volunteer and past director of its Southeast District, appeared in a March, 2012 article. In it, Mr. Tolbert expounds on seven cautions to help avoid costly mistakes, from overstocking of SKUs and excessive (and costly) inventories, to missed sales opportunities and missed customer due dates.
Briefly recapping the key points…
1. Using shipment history: Sales forecasting systems use sales history data to generate statistical forecasts for future periods. This can wreak havoc if, for example, your customers place large orders in July that are backordered until September, and then shipped. Do you recognize the sale as being in July or September? Probably September. Thus, your forecast will end up perpetuating demand in the wrong period.
2. Relying on bad data: Even if you avoided #1, a one-time or non-recurring sale can still skew the data and thus your future forecasting horizon. Special sales promotions, pipeline order fills and spikes in demand can easily cause these hiccups.
3. Excessive “gut feel” overrides: Companies plan, then override by gut feel quite often. A red flag should arise if those guts are overriding more than 10 or 20% of your plan. Adjust the future only if you know something about the future that is not reflected in the past data for forecasting.
4. Poor event planning: Sometimes, this is the opposite of no. 3. New item introductions, trade show specials and item substitutions can all affect planning. They are not reflected in past sales and thus may affect your plan. Frequent plan reviews help ameliorate the ‘sins’ of no’s. 3 and 4.
5. Senior management meddling: Inventory and service levels are best left to the planning experts in your supply chain. Unfortunately, this is the easiest mistake to recognize, and often the hardest to correct.
6. Failure to measure sales forecast accuracy: You have to measure. Measure forecasts at the SKU and product family levels. You have to account for occasional long lead times. You have to determine whether the statistical forecast or the planners’ overrides are more accurate over time. And you have to look for ‘bias’ in your forecasts, resulting in forecasts that are consistently too high or too low.
7. Safety stock based on forecast error. Safety stock exists for those times when demand exceeds forecast. But the traditional safety stock calculations (often based on a ‘service rate,’ like, say 98% order-fill) don’t distinguish between periods when a forecast is too high or too low. Strategies including statistical modeling techniques that eliminate bias, or inventory strategies based on safety time rather than safety stock calculations that may include forecasting errors, can help.